mcclowes/ai-aware-code-structure
Use when deciding how to organise code in an AI-assisted codebase — whether to split or merge a file, where to draw module boundaries, how big a file should get, whether to separate logic from rendering/styles/data, whether to use barrel/index re-export files, or where types and test files should live. Triggers on "should I split this file", "this file is getting too big", "separate concerns", "where should this code live", "are barrel files worth it", "central types file or co-locate", "colocate tests or a tests folder", reviewing or refactoring file/module organisation, structuring a new component or feature, or any architecture decision where part of the audience is an AI coding agent. Apply this whenever someone is choosing how to lay code out across files and an LLM will be reading or editing it, even if they only say "refactor this" or "clean up the structure" without mentioning AI.
83
87%
Does it follow best practices?
Impact
82%
1.54xAverage score across 9 eval scenarios
Passed
No known issues
criteria.jsonevals/scenario-5/
{
"context": "A 420-line UserService bundles auth, persistence, email, and analytics, defended as 'everything in one place for the agent'. The criteria check whether the answer splits by responsibility on clean seams and pushes back on the misapplied rationale.",
"type": "weighted_checklist",
"checklist": [
{
"name": "Recommends splitting by responsibility",
"description": "Recommends splitting the class by responsibility and does NOT endorse keeping all four concerns in one file",
"max_score": 8
},
{
"name": "Identifies independent seams",
"description": "Identifies the seams (auth / persistence / email / analytics) as genuinely independent and isolatable",
"max_score": 6
},
{
"name": "Ceiling and/or unrelated concerns",
"description": "Cites the ~300-line ceiling AND/OR multiple-unrelated-concerns as the reason, not blind line count",
"max_score": 6
},
{
"name": "Pushes back on keep-together rationale",
"description": "Pushes back on the 'keep everything in one place for the agent' rationale as misapplied to a god object",
"max_score": 6
}
]
}